Transcribing voiced musical notes for creating, practicing and sharing of musical harmonies

Abstract

Method for transcription of voiced musical note includes segmenting an electronic audio signal into a plurality of musical note segments, and sampling each note segment to obtain a plurality of audio samples. For each audio sample, an autocorrelation is computed to determine a probability value associated with certain audio frequencies contained within the audio sample. Local maxima are identified in the energy associated with one or more audio frequencies comprising each audio sample and a corrective function applied to reduce the occurrence of octave errors. A true pitch of each note segment is then determined by converting the pitch identification problem to one involving a shortest path through a graph or network of node. Edge weights are computed for a plurality of adjacent nodes i, j comprising the graph, where each node represents one of the musical note segments.